Bayesian Analyses of Nonhomogeneous Autoregressive Processes
Abstract
This paper considers nonhomogeneous autoregressive processes which are special cases of the vector-valued autoregressive processes considered by Anderson (1978) for the analysis of panel survey data. The authors point out that, for a nonhomogeneous autoregressive process of order higher than one, the least-squares estimates cannot be obtained unless repeated measurements are made on the time series. Presented are two Bayesian approaches based on Kalman filter models which alleviate the above difficulty and result in an alternative strategy for the analyses of nonhomogeneous autoregressive processes. In the first approach the notion of exchangeability plays a key role, whereas for the second approach, which results in an adaptive Kalman filter model, an approximation due to Lindley facilitates the necessary computations for inference.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1986
- Accession Number
- ADA175725
Entities
People
- Nozer D. Singpurwalla
- Refik Soyer
- Theodore W. Anderson
Organizations
- Stanford University